The first, “restrictive” approach consisted of the mutagenesis of the pAB24 plasmid, which contains the wild-type parD operon, and the consecutive screening for Kid mutants unable to corepress the parD promoter. A possible limitation of this approach was that Kid mutations that affected the interactions between Kis and Kid required for the neutralization of Kid toxicity should lead to growth inhibition and therefore could be counterselected. The second, “permissive” approach relieved this constraint: instead of pAB24, we used pB24, a mutated version of pAB24 carrying the kid85 mutation. This mutation abolishes the toxicity of the Kid protein but maintains its activity as a corepressor (13), thereby permitting the eventual isolation of kid mutations which lead to derepression by disrupting the interactions between Kis and Kid.

Thus, the pB24 or pAB24 plasmid was mutagenized in vitro with hydroxylamine as described previously (6) and this DNA was used to transform the Escherichia coli CSH16 strain (Δlac supE) containing pOM34, a mini-R1 recombinant in which the lacZ gene is transcriptionally fused to the parD promoter (9). In this background, parD mutants that failed to repress in trans the parD promoter elicited β-galactosidase synthesis and led to the formation of red transformants on MacConkey agar plates supplemented with lactose (Fig. 1A). Plasmids from red colonies were isolated and retransformed in the same strain, and those reproducing the mutant phenotype were kept for further analysis. The sequences of the complete parD operon in each of the mutant plasmids were determined and led to the identification of 11 different mutations in the kid gene (Fig. 1B). As expected from the mutagenic action of hydroxylamine, the changes isolated were C-G/A-T transitions and introduced missense or nonsense mutations. The latter led to truncated proteins that were not toxic (data not shown), in agreement with previous data that showed the relevance of the carboxyl end of the protein in toxicity (4, 12).

FIG. 1. Kid mutations that affect the autoregulation of the parD system. (A) CSH16 cells carrying the reporter of parD expression plasmid pOM34 and the mutated derivatives of plasmid pB24 (Tetr kis74 kid85) or pAB24 (Tetr kis+ kid+) were streaked (more ...)

Therefore, we concentrated our analysis on the more informative missense mutations. kid4 (which leads to a G4E amino acid change in Kid), kid29 (T29I), kid70 (G70D), kid74 (C74Y), kid79 (T79A), kid87 (G87R), kid91 (E91K), and kid94 (P94L). With the exception of G4E and T29I, the changes were clustered at the carboxy-terminal third of the Kid protein in a 20-amino-acid region bounded by residues 74 to 94. In a previous search for nontoxic mutants of Kid, the changes T29I, P94L, and G70S (the last affecting the same residue as G70D of our current study) were shown to affect both the toxic and the autoregulatory activities of Kid (13). Therefore, their isolation in our present approach was theoretically expected, as internal controls of nonspecific disruption of the autoregulatory activity of Kid. These three mutations will not be considered further in this work. Among the remaining mutations, we noticed that the C74Y, T79A, and G87R changes were isolated in the permissive approach, using pB24 (Fig. 1B). This suggested either that the observed deregulation phenotype could be the consequence of a synergistic effect between these mutations and the mutations originally present in pB24 (kis74 and kid85) or that the nontoxic kid85 mutation present in pB24 could prevent an eventual negative effect on growth due to the additional mutations.

To test these hypotheses, each of the kid74, kid79, and kid87 mutations was introduced separately in pAB24 by using a site-directed mutagenesis kit (Promega). The resultant plasmids were used to transform the MLM373 strain [Δ(lac, pro) supE thi)] (13), which also contained the pMLM132 plasmid, a mini-F derivative that bears the lacZ gene under the control of the parD promoter (13). parD expression levels were monitored in β-galactosidase assays that were carried out as described previously (8). The Kid changes T79A and G87R did not significantly affect the autoregulatory activity of the protein (Fig. 2A), confirming that the deregulation phenotypes associated with the kid79 and kid87 mutations were due to interactions with the kid85 or kis74 mutation present in pB24. In contrast, the Kid C74Y change provoked a fivefold increase in parD promoter activity (Fig. 2A), confirming that the single C74Y change was sufficient to lead to deregulation. Furthermore, growth was severely inhibited in cells that contained the pAB24 plasmid bearing the kid74 mutation but not in those that contained the other mutant or wild-type pAB24 plasmid (Fig. 2B). Growth inhibition was not observed when the kid74 mutation was present in pB24 (data not shown; see also Fig. 1A). Therefore, this indicates that the C74Y change has been counterselected by the pAB24 restrictive approach and that the negative effect of a Kid C74Y change is suppressed by the mutations present in pB24. In addition, these experiments confirmed that the kid4(G4E) and kid91(E91K) mutations lead to deregulation of parD expression, as revealed by the respective six- and threefold increases in β-galactosidase activity (Fig. 2A).

FIG. 2. Effect of mutations on parD autoregulation and cell growth. (A) Autoregulation. MLM373 cells containing the parDp-lacZ fusion carrier plasmid pMLM132 and the derivatives of the pAB24 plasmid carrying the different mutations were grown to mid-exponential (more ...)

To determine if the mutations kid4(G4E), kid74(C74Y), and kid91(E91K), which affected Kid coregulatory functions, would also influence its toxic activity, the kid genes carrying single mutations were isolated and cloned under the control of a T7 RNA polymerase-dependent promoter (15) in the pET3d-his vector (Stratagene) to yield the series of pET3d-hiskid mutant plasmids. As a control, the “nontoxic” kid85 mutant allele was also included in the analysis. Expression of the cloned genes was induced in the BL21(DE3) strain (Stratagene) by adding isopropyl-β-d-thiogalactopyranoside (IPTG) to the medium. In addition, the strains contained the pMLM126 plasmid, a replication-thermosensitive pSC101 derivative in which the kis antitoxin gene was placed under the control of the arabinose-inducible araBAD promoter. Following induction with IPTG, inhibition of cell growth was observed in cells in which the wild-type kid gene was expressed but not in those that expressed the kid85 mutant, as expected (Fig. 3, compare panels B and C). Moreover, the three deregulated kid mutants tested (producing Kid mutants G4E, C74Y, and E91K) showed a toxic phenotype (Fig. 3B and C). The phenotype was comparable to the one displayed by the wild-type Kid protein in the case of G4E but slightly less marked in the case of C74Y and significantly more severe in the case of E91K (Fig. 3C). In fact, background levels of E91K protein obtained in the absence of IPTG induction were sufficient to dramatically affect cell viability (Fig. 3B). When overproduction of the Kis antitoxin was induced with arabinose to neutralize the toxic effect of Kid, a neat recovery of viability was observed in cells that produced either the wild-type Kid protein or the mutant protein G4E or C74Y (Fig. 3, compare panels C and D). Under the same conditions, the E91K mutant protein was resistant to neutralization by Kis. However, neutralization of E91K toxicity was observed when Kis was overproduced in the absence of kid induction (Fig. 3, compare panels A and B). This suggested that the hypertoxic phenotype of the E91K mutant needed a large excess of Kis antitoxin cellular levels to be neutralized. This is consistent with the isolation of the kid91(E91K) mutation in the pAB24 plasmid, because the posttranscriptional control of parD expression ensures that the antitoxin is produced at higher levels than the toxin (2, 10).

FIG. 3. Analysis of toxicity of the Kid mutants and neutralization by the Kis antitoxin. BL21(DE3) cells carrying the kis overproducer plasmid pMLM126 and the different pET3d-hiskid wild-type or mutant plasmids were grown exponentially to an A600 of 0.4 in Luria (more ...)

To summarize, we found three single-amino-acid changes in the Kid toxin, G4E, C74Y, and E91K, that were able to affect the coregulatory activity of Kid without causing a loss of the toxic activity of the protein. The G4E change severely impairs the coregulation activity but maintains both the toxicity of the protein and its capacity to interact with the Kis antitoxin to form a nontoxic complex. Therefore, this change may affect Kid-Kis interactions in a way that the complex is proficient in neutralization of toxicity but deficient in properly interacting with the parD operator sequences and/or host factors required for parD regulation. A future biochemical characterization of this mutant protein will help to address this issue. Likewise, the Kid C74Y protein is efficiently neutralized by a supply of Kis in a trans configuration. However, introducing the kid74(C74Y) mutation in a wild-type parD operon leads to deregulation of parD expression and severe cell growth inhibition, suggesting that the Kis-Kid C74Y complex may assemble in a first step and then be altered, upon binding to the parD operator sequences or to other proteins of the regulatory complex, in such a way that autoregulation is impaired and unneutralized toxic Kid C74Y protein becomes exposed to its target.

This hypothesis is supported by the recent structural data on the closely related MazE-MazF TA complex, which argues in favor of a rearrangement of the complex upon DNA binding (7). Strikingly, the C74 residue is bordered by two residues, R73 and D75, that are essential for the toxic activity of Kid (13). Moreover, the region of the homologous MazF toxin that corresponds to the R73-D75 region of Kid is involved in contacts with the carboxy-terminal region of the MazE antitoxin, suggesting that the C74 residue might be crucial for Kis-Kid interactions required for neutralization as well as for autoregulation. Finally, the E91K mutant protein is neutralized by the Kis antitoxin, provided that the latter is present in excess. Perhaps the most striking feature of this mutant is that, compared with the wild-type toxin, its dramatic cytotoxic effect requires very low levels of expression. It might be that the E91K change favors a more efficient interaction of the Kid toxin with its target, in a manner that is less efficiently competed by the Kis antitoxin. Support for this hypothesis has been provided by the structural analyses of the Kid toxin and of the MazE-MazF complexes, which suggest that the target may compete with the C-terminal region of the antitoxin to bind to the toxin (7, 4). Addressing this issue awaits further structural and functional information on toxin-target complexes, as well as on autoregulation complexes bound to DNA.

This research was supported by grants from the European Union (grant QLK2-CT-2000-00634), the Ministerio de Educación y Cultura, Spain (grant BIO99-0859-CO3-01), the “Programa de Grupos Estratégicos de la Comunidad de Madrid,” 2000-2003, and by the Spanish REIP Network of the “Fondo de Investigaciones Sanitarias.”